Skip to content

feat: context-updater agent, token budget compliance, headless mode docs#6

Merged
Nathan Schram (nathanschram) merged 2 commits intomainfrom
feat/v1.2-context-updater-token-budgets
Mar 11, 2026
Merged

feat: context-updater agent, token budget compliance, headless mode docs#6
Nathan Schram (nathanschram) merged 2 commits intomainfrom
feat/v1.2-context-updater-token-budgets

Conversation

@nathanschram
Copy link
Member

Summary

  • context-updater agent — autonomous agent launched by hooks to update stale AI context files after structural changes, closing the detection-to-action loop
  • Token budget compliance — trimmed all 3 skills and 2 rules to within budgets (was 5 warnings, now 0)
  • Headless mode documentation — documented claude -p limitation (Skill auto-triggering: recall=0% in headless mode (claude -p) regardless of description content anthropics/claude-code#32184) in troubleshooting, CHANGELOG, CLAUDE.md
  • Activation eval results — 3 eval runs + description optimisation findings (300 queries) confirming platform limitation
  • Hook updates — all hooks now reference context-updater agent instead of suggesting manual commands
  • Typos config — excluded banned phrases test from spell check (contains intentional US English)

Token Budget Results

File Before After Budget
ai-context SKILL.md ~4010 tokens ~1179 tokens 2000
context-guard SKILL.md ~2693 tokens ~1274 tokens 2000
context-verify SKILL.md ~2083 tokens ~826 tokens 2000
context-awareness rule ~651 tokens ~402 tokens 500
context-quality rule ~778 tokens ~405 tokens 500

Test plan

  • Hook unit tests (65/65 pass)
  • Banned phrases check (clean)
  • Token budget validation (0 warnings)
  • llms.txt validation (0 errors, 0 warnings)
  • Frontmatter validation (all valid)
  • Spell check / typos (clean)
  • Markdownlint (0 errors, 22 files)

🤖 Generated with Claude Code

…document headless mode limitations

- Add context-updater agent for autonomous context file maintenance
- Hooks now reference context-updater agent instead of suggesting manual commands
- Trim all 3 skills and 2 rules to within token budgets (was 5 warnings, now 0)
- Document claude -p headless mode limitation (anthropics/claude-code#32184)
- Add activation eval results (3 runs) and description optimisation findings
- Update troubleshooting guide with affected/unaffected scenarios
- Exclude banned phrases test from typos spell check (intentional US English)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@chatgpt-codex-connector
Copy link

You have reached your Codex usage limits for code reviews. You can see your limits in the Codex usage dashboard.
To continue using code reviews, you can upgrade your account or add credits to your account and enable them for code reviews in your settings.

…links

- Replace `ls` with `find` in activation-evals.yml and docs-ci.yml (SC2012)
- Group GITHUB_OUTPUT redirects in activation-evals.yml (SC2129)
- Fix Keep a Changelog URL version (1.1.1 → 1.1.0)
- Exclude release tag URLs from lychee link checker (tags created by release-please)

Co-Authored-By: Claude Opus 4.6 <noreply@anthropic.com>
@nathanschram Nathan Schram (nathanschram) merged commit a4b6159 into main Mar 11, 2026
9 checks passed
@nathanschram Nathan Schram (nathanschram) deleted the feat/v1.2-context-updater-token-budgets branch March 11, 2026 05:32
Sign up for free to join this conversation on GitHub. Already have an account? Sign in to comment

Labels

None yet

Projects

None yet

Development

Successfully merging this pull request may close these issues.

1 participant